Stochastic logarithm explained

In stochastic calculus, stochastic logarithm of a semimartingale

Y

such that

Y ≠ 0

and

Y- ≠ 0

is the semimartingale

X

given by[1] dX_t=\frac,\quad X_0=0.In layperson's terms, stochastic logarithm of

Y

measures the cumulative percentage change in

Y

.

Notation and terminology

The process

X

obtained above is commonly denoted

l{L}(Y)

. The terminology stochastic logarithm arises from the similarity of

l{L}(Y)

to the natural logarithm

log(Y)

: If

Y

is absolutely continuous with respect to time and

Y0

, then

X

solves, path-by-path, the differential equation \frac = \frac,whose solution is

X=log|Y|-log|Y0|

.

General formula and special cases

Y

(other than

Y0,Y- ≠ 0

), one has\mathcal(Y)_t = \log\Biggl|\frac\Biggl|+\frac12\int_0^t\frac+\sum_\Biggl(\log\Biggl| 1 + \frac \Biggr|-\frac\Biggr),\qquad t\ge0,where

[Y]c

is the continuous part of quadratic variation of

Y

and the sum extends over the (countably many) jumps of

Y

up to time

t

.

Y

is continuous, then \mathcal(Y)_t = \log\Biggl|\frac\Biggl|+\frac12\int_0^t\frac,\qquad t\ge0.In particular, if

Y

is a geometric Brownian motion, then

X

is a Brownian motion with a constant drift rate.

Y

is continuous and of finite variation, then\mathcal(Y) = \log\Biggl|\frac\Biggl|.Here

Y

need not be differentiable with respect to time; for example,

Y

can equal 1 plus the Cantor function.

Properties

\DeltaX-1

, then

l{L}(l{E}(X))=X-X0

. Conversely, if

Y0

and

Y- ≠ 0

, then

l{E}(l{L}(Y))=Y/Y0

.

log(Yt)

, which depends only of the value of

Y

at time

t

, the stochastic logarithm

l{L}(Y)t

depends not only on

Yt

but on the whole history of

Y

in the time interval

[0,t]

. For this reason one must write

l{L}(Y)t

and not

l{L}(Yt)

.

Y

.

Y

that are absorbed in zero after jumping to zero. Such definition is meaningful up to the first time that

Y

reaches

0

continuously.[2]

Useful identities

Y(1),Y(2)

do not vanish together with their left limits, then\mathcal\bigl(Y^Y^\bigr) = \mathcal\bigl(Y^\bigr) + \mathcal\bigl(Y^\bigr) + \bigl[\mathcal{L}\bigl(Y^{(1)}\bigr),\mathcal{L}\bigl(Y^{(2)}\bigr)\bigr].

1/l{E}(X)

:[3] If

\DeltaX-1

, then\mathcal\biggl(\frac\biggr)_t = X_0-X_t-[X]^c_t+\sum_\frac.

Applications

Q

be a probability measure equivalent to another probability measure

P

. Denote by

Z

the uniformly integrable martingale closed by

Zinfty=dQ/dP

. For a semimartingale

U

the following are equivalent:

U

is special under

Q

.

U+[U,l{L}(Z)]

is special under

P

.

Q

-drift of

U

equals the

P

-drift of

U+[U,l{L}(Z)]

.

See also

Notes and References

  1. Book: Jacod. Jean. Limit theorems for stochastic processes. Shiryaev. Albert Nikolaevich. 2003. Springer. 3-540-43932-3. 2nd. Berlin. 134–138. 50554399.
  2. Larsson. Martin. Ruf. Johannes. 2019. Stochastic exponentials and logarithms on stochastic intervals — A survey. Journal of Mathematical Analysis and Applications. en. 476. 1. 2–12. 10.1016/j.jmaa.2018.11.040. 119148331 . free. 1702.03573.
  3. Larsson. Martin. Ruf. Johannes. 2019. Stochastic exponentials and logarithms on stochastic intervals — A survey. Journal of Mathematical Analysis and Applications. en. 476. 1. 2–12. 10.1016/j.jmaa.2018.11.040. 119148331 . free. 1702.03573.